Generate ξ m,i for each direction i given H, σ 1 and m (Eq. 2) calculate X’ m,i for each direction i (Eq. 1) given ξ m,i and X m, which corresponds for.

Slides:



Advertisements
Similar presentations
Medium-range Ensemble Streamflow forecast over France F. Rousset-Regimbeau (1), J. Noilhan (2), G. Thirel (2), E. Martin (2) and F. Habets (3) 1 : Direction.
Advertisements

Poster template by ResearchPosters.co.za Effect of Topography in Satellite Rainfall Estimation Errors: Observational Evidence across Contrasting Elevation.
Sampling: Final and Initial Sample Size Determination
Scaling Laws, Scale Invariance, and Climate Prediction
Details for Today: DATE:3 rd February 2005 BY:Mark Cresswell FOLLOWED BY:Assignment 2 briefing Evaluation of Model Performance 69EG3137 – Impacts & Models.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Improving COSMO-LEPS forecasts of extreme events with.
Gridded OCF Probabilistic Forecasting For Australia For more information please contact © Commonwealth of Australia 2011 Shaun Cooper.
Estimation of Rainfall Areal Reduction Factors Using NEXRAD Data Francisco Olivera, Janghwoan Choi and Dongkyun Kim Texas A&M University – Department of.
Landslide Susceptibility Mapping to Inform Land-use Management Decisions in an Altered Climate Muhammad Barik and Jennifer Adam Washington State University,
Improving Probabilistic Ensemble Forecasts of Convection through the Application of QPF-POP Relationships Christopher J. Schaffer 1 William A. Gallus Jr.
PROVIDING DISTRIBUTED FORECASTS OF PRECIPITATION USING A STATISTICAL NOWCAST SCHEME Neil I. Fox and Chris K. Wikle University of Missouri- Columbia.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Quantitative precipitation forecasts in the Alps – first.
CS 8751 ML & KDDEvaluating Hypotheses1 Sample error, true error Confidence intervals for observed hypothesis error Estimators Binomial distribution, Normal.
Chapter 1 Ways of Seeing. Ways of Seeing the Atmosphere The behavior of the atmosphere is very complex. Different ways of displaying the characteristics.
Introduction to Educational Statistics
Providing distributed forecasts of precipitation using a Bayesian nowcast scheme Neil I. Fox & Chris K. Wikle University of Missouri - Columbia.
Statistics and Probability Theory Prof. Dr. Michael Havbro Faber
MOS Performance MOS significantly improves on the skill of model output. National Weather Service verification statistics have shown a narrowing gap between.
Ensemble Post-Processing and it’s Potential Benefits for the Operational Forecaster Michael Erickson and Brian A. Colle School of Marine and Atmospheric.
Chapter 13 – Weather Analysis and Forecasting. The National Weather Service The National Weather Service (NWS) is responsible for forecasts several times.
Verification has been undertaken for the 3 month Summer period (30/05/12 – 06/09/12) using forecasts and observations at all 205 UK civil and defence aerodromes.
SRNWP workshop - Bologne Short range ensemble forecasting at Météo-France status and plans J. Nicolau, Météo-France.
1 GOES-R AWG Hydrology Algorithm Team: Rainfall Probability June 14, 2011 Presented By: Bob Kuligowski NOAA/NESDIS/STAR.
Results of the WMO Laboratory Intercomparison of rainfall intensity gauges Luca G. Lanza University of Genoa WMO (Project Leader) DIAM UNIGE September.
ESA DA Projects Progress Meeting 2University of Reading Advanced Data Assimilation Methods WP2.1 Perform (ensemble) experiments to quantify model errors.
Two Sample Tests Nutan S. Mishra Department of Mathematics and Statistics University of South Alabama.
Short-Range Ensemble Prediction System at INM José A. García-Moya & Carlos Santos SMNT – INM COSMO Meeting Zurich, September 2005.
How can LAMEPS * help you to make a better forecast for extreme weather Henrik Feddersen, DMI * LAMEPS =Limited-Area Model Ensemble Prediction.
Chapter 6 Lecture 3 Sections: 6.4 – 6.5.
STEPS: An empirical treatment of forecast uncertainty Alan Seed BMRC Weather Forecasting Group.
Improving Ensemble QPF in NMC Dr. Dai Kan National Meteorological Center of China (NMC) International Training Course for Weather Forecasters 11/1, 2012,
Celeste Saulo and Juan Ruiz CIMA (CONICET/UBA) – DCAO (FCEN –UBA)
Latest results in verification over Poland Katarzyna Starosta, Joanna Linkowska Institute of Meteorology and Water Management, Warsaw 9th COSMO General.
Simulations of present climate temperature and precipitation episodes for the Iberian Peninsula M.J. Carvalho, P. Melo-Gonçalves and A. Rocha CESAM and.
Determination of Sample Size: A Review of Statistical Theory
Short-Range Ensemble Prediction System at INM José A. García-Moya SMNT – INM 27th EWGLAM & 12th SRNWP Meetings Ljubljana, October 2005.
PREDICTABILITY OF WESTERN NORTH PACIFIC TROPICAL CYCLONE EVENTS ON INTRASEASONAL TIMESCALES WITH THE ECMWF MONTHLY FORECAST MODEL Russell L. Elsberry and.
Feng Zhang and Aris Georgakakos School of Civil and Environmental Engineering, Georgia Institute of Technology Sample of Chart Subheading Goes Here Comparing.
PROBABILITY AND STATISTICS FOR ENGINEERING Hossein Sameti Department of Computer Engineering Sharif University of Technology Mean, Variance, Moments and.
Refinement and Evaluation of Automated High-Resolution Ensemble-Based Hazard Detection Guidance Tools for Transition to NWS Operations Kick off JNTP project.
Dept of Bioenvironmental Systems Engineering National Taiwan University Lab for Remote Sensing Hydrology and Spatial Modeling STATISTICS Linear Statistical.
Typhoon Forecasting and QPF Technique Development in CWB Kuo-Chen Lu Central Weather Bureau.
Chapter 6 Lecture 3 Sections: 6.4 – 6.5. Sampling Distributions and Estimators What we want to do is find out the sampling distribution of a statistic.
The uncertainty in the prediction of flash floods in the Northern Mediterranean environment; single site approach and multi-catchment system approach CIMA.
STOCHASTIC HYDROLOGY Stochastic Simulation of Bivariate Distributions Professor Ke-Sheng Cheng Department of Bioenvironmental Systems Engineering National.
Lab for Remote Sensing Hydrology and Spatial Modeling Dept of Bioenvironmental Systems Engineering National Taiwan University 1/45 GEOSTATISTICS INTRODUCTION.
Verification of ensemble systems Chiara Marsigli ARPA-SIMC.
Nathalie Voisin 1, Florian Pappenberger 2, Dennis Lettenmaier 1, Roberto Buizza 2, and John Schaake 3 1 University of Washington 2 ECMWF 3 National Weather.
Stochastic Hydrology Random Field Simulation Professor Ke-Sheng Cheng Department of Bioenvironmental Systems Engineering National Taiwan University.
CHAPTER 2.3 PROBABILITY DISTRIBUTIONS. 2.3 GAUSSIAN OR NORMAL ERROR DISTRIBUTION  The Gaussian distribution is an approximation to the binomial distribution.
NCAR, 15 April Fuzzy verification of fake cases Beth Ebert Center for Australian Weather and Climate Research Bureau of Meteorology.
DOWNSCALING GLOBAL MEDIUM RANGE METEOROLOGICAL PREDICTIONS FOR FLOOD PREDICTION Nathalie Voisin, Andy W. Wood, Dennis P. Lettenmaier University of Washington,
VERIFICATION OF A DOWNSCALING SEQUENCE APPLIED TO MEDIUM RANGE METEOROLOGICAL PREDICTIONS FOR GLOBAL FLOOD PREDICTION Nathalie Voisin, Andy W. Wood and.
SUMMARY EQT 271 MADAM SITI AISYAH ZAKARIA SEMESTER /2015.
Chapter 7: The Distribution of Sample Means
LESSON 5 - STATISTICS & RESEARCH STATISTICS – USE OF MATH TO ORGANIZE, SUMMARIZE, AND INTERPRET DATA.
11 Short-Range QPF for Flash Flood Prediction and Small Basin Forecasts Prediction Forecasts David Kitzmiller, Yu Zhang, Wanru Wu, Shaorong Wu, Feng Ding.
Improving Numerical Weather Prediction Using Analog Ensemble Presentation by: Mehdi Shahriari Advisor: Guido Cervone.
Performance assessment of a Bayesian Forecasting System (BFS) for realtime flood forecasting Biondi D. , De Luca D.L. Laboratory of Cartography and Hydrogeological.
LEPS VERIFICATION ON MAP CASES
Verifying and interpreting ensemble products
Precipitation Products Statistical Techniques
Methodology to integrate dynamical and statistical weather forecasts
Nathalie Voisin, Andy W. Wood and Dennis P. Lettenmaier
Stochastic Hydrology Random Field Simulation
N. Voisin, J.C. Schaake and D.P. Lettenmaier
Christoph Gebhardt, Zied Ben Bouallègue, Michael Buchhold
Verification of Tropical Cyclone Forecasts
Some Verification Highlights and Issues in Precipitation Verification
Short Range Ensemble Prediction System Verification over Greece
Presentation transcript:

generate ξ m,i for each direction i given H, σ 1 and m (Eq. 2) calculate X’ m,i for each direction i (Eq. 1) given ξ m,i and X m, which corresponds for the first step of this iterative disaggregating process to a EPS forecasted rainfall amount. calculate the four disaggregated R k=1:4 rainfall amounts given the three X’ m,i fluctuations and the initial X m amount (see figure D: a simple four unknown variables and four equations system). E. Gaborit 1, F. Anctil 1, G. Pelletier 1, and V. Fortin 2 [1] Civil and Water Engineering Department, Laval University, Quebec, Canada. Contact: [2] Numerical Weather Prediction Research, Environment Canada, Dorval, Canada On the reliability of spatially disaggregated global ensemble rainfall forecasts 1. Problem statement B1- Ensemble Prediction System (EPS) member’s forecasted rainfall field example Accumulated precipitation over 3 hours* amount (mm) A- Studied watershed (500 km 2 ) 100 km 70 km B2- Possible actual rainfall field at a 6 km resolution There is a gap between the resolution at which ensemble rainfall forecasts are currently available and the small scale of watersheds sometimes used in hydrologic studies: an average precipitation amount forecasted for a 7000 km 2 sector (EPS resolution, see figure B1) often hides a strong rainfall variability inside that sector, such that local rainfall observations will actually exhibit areas experiencing no rainfall and others with amounts much higher than the average value forecasted for the global sector (difference between B1 and B2).  Ensemble forecasts are of potentially high interest for decision-making in hydrologic studies  Their resolution currently limits their use for studies conducted on small watersheds. 2. Objective Bridge the spatial gap between ensemble rainfall forecasts’ original scale and small watersheds by increasing the rainfall variance inside each initial pixel of an EPS product member rainfall field.  Independently spatially disaggregate each 21 members (i.e. scenarios) of the ensemble product  Preserve the original mean rainfall amount inside each initial pixel 3. Methodology Resolution of the “LAM” deterministic model: 2.5 x 2.5 km There are at least three options for downscaling low-resolution ensemble forecasts: statistical downscaling, dynamical downscaling, and stochastic downscaling. Different disaggregation approaches based on a method proposed by Périca and Foufoula-Georgiou (1996) have been implemented to disaggregate the original Environment Canada’s EPS down to a 6 km resolution. This method consists in a recursive stochastic disaggregation process based on scaling relationships. It belongs to stochastic downscaling. In order not to modify the given original ensemble forecast, the choice has been made to:  3.1 The method’s theory: * 3 hours is the time resolution of the used EPS Rainfall directional fluctuations: R1R1 R2R2 R3R3 R4R Standardized fluctuations: initial pixel average rainfall amount = X m Downscaled pixel rainfall amount = R k=1:4 D. Definition of standardized rainfall fluctuationsE. Scaling relationship Over an area of for example km 2, standardized rainfall fluctuations of any direction follow the same normal distribution law (Eq. 2) with zero mean and whose standard deviation shows a simple behavior over relative scales m. Standardized rainfall fluctuations Standard deviation at relative scale m Parameters of the method relative scale m01234 corresponding resolution (km) F. Method algorithm: each iteration increases the resolution by 2G. Different disaggregation approaches For each pixel of a given EPS member’s rainfall forecast field at a relative scale m (start from m corresponding to the initial EPS resolution): (Eq. 2) (Eq. 1) m=m-1 Different ways to calculate ξ m for a pixel have led to different approaches : Use the mean H and σ 1 values found by the authors presenting the method (APPROACH E1). σ m,i is a function of scale m, direction i and pixel’s rainfall amount (APPROACH E2). ξ m,i is calculated based on the “LAM” deterministic product forecasted rainfall fields (APPROACH E3) or based on all available deterministic products (approach E4).  3.2 Evaluation: direct confrontation with observed rainfall amounts Note: approaches E1 and E2 involving values’ selection from a probability density function, 5 repetitions of each of these approaches have been implemented to study the potential subsequent differences in the final disaggregated rainfall fields.  Scores: all scores are calculated for one product considering all days, emission hours, horizons and pixels simultaneously. The deterministic product (D-T) used is the one whose resolution is just under the considered one (for resolutions 12 and 6km it is the “LAM” itself) Deterministic evaluation (using the ensembles’ mean or deterministic products): Probabilistic evaluation (using the ensembles): Mean absolute error (MAE): directly comparable to the CRPS. Relative operating characteristics score (ROC score): see Peterson et al., Scores based on contingency tables with different rainfall thresholds values : see Rezacova et al Continuous Ranked Probability score (CRPS, see Matheson and Winkler, 1976): comparable to the MAE. ROC score (see above). Talagrand diagrams (see Olsson and Lindström, 2008). Reliability diagrams (see Olsson and Lindström, 2008).  Period: 9 consecutive days of summer 2009 with strong convective events  Observed data: each pixel’s forecasted amount is compared with the mean observed amount of Quebec City rain gages (see figure A) located inside it. The number of pixels with observed data depends on the considered resolution. H. Example ensemble forecasts and observations on a 50-km resolution pixel, for one day and one emission hour Time since emission hour (hours) Observations Ensemble forecasts Rainfall amount over 3 hours (mm) Forecasts/ Observations pairs used in a score calculation Δt=3h Emission hour (2 per day) Maximum considered horizon in the evaluation (72 h) 4. Results J. Symbols of the different evaluated products bilinear interpolation of the original EPS (E-B) sub-pixels inside an original EPS one have the same rainfall amount as the latter (E-0) disaggregated products using the method proposed by the authors and using four different approaches (see point G. on the left). “LAM” deterministic product aggregated to the considered resolution (D-L) rep 1:5 repetitions 1 to 5 (approaches E-1 and E-2) E-1 E-2 E-3 E-4 Target for score (value for perfect forecasts) K. MAE at a 06-km resolution using the mean of the ensembles and deterministic products MAE (mm) Complete data Forecasts>0Forecasts=0 Legend (see J.)  Better overall performance of the “LAM” product compared to the mean of the disaggregated ensembles Target value K. MAE vs CRPS at resolution 50km and complete data MAECRPS Deterministic product / mean of the ensembles Ensemble products M. ROC Score at resolution 50km and rainfall threshold value 0.05mm Deterministic product / mean of the ensembles Ensemble products Target Value =0 Target Value =1 N. ROC Score for (mean) rainfall threshold value 0.05 mm, as a function of the resolution ensemble and deterministic products only Target Value =1 ensemble and deterministic products only L. CRPS as a function of the resolution Target Value =0  Taking the ensemble products leads to better results than taking their deterministic counterparts  The overall quality of the forecasts is preserved through scales (products E-B, E-0, E-1 and E-2)  The variance-enhanced products are of similar quality than the bi-linearily interpolated one O. Talagrand diagrams at resolution 6km without cases with no rainfall for all the ensemble’s members Product E-BProduct E-1, repetition 1 Target: flat shape P. Reliability diagrams at resolution 6km, for mean rainfall threshold 1 mm Product E-1, repetition 1Product E-B Target line Q. Variance of rainfall amounts over the entire disaggregated grid as a function of scale R. Rainfall amounts’ variance box-plot for pixels with observed values at resolution 6km Observations Product 95% boundary 75% boundary median 25% boundary 5% boundary 5 Observation’s rank n (n=1:22) 1 Y axis: mean number of cases with observation in rank n Observation’s rank n (n=1:22)  Better variance and dispersion for the disaggregated products using the method proposed by Périca and Foufoula-Gergiou (1996) than for the one using bilinear interpolation See J. Y axis: mean effective probability X axis: nominal probability Median and 95% confidence interval shown 5. Conclusion This work shows that (the simple approach E-1 of) the method proposed by Périca and Foufoula- Georgiou (1996) represents an advantage over simpler methods to spatially disaggregate Ensemble Forecasts, since it allows increasing (thus improving) the rainfall variability inside an initial EPS member’s pixel while leading to similar performances from other scores’ point of views. However, the method currently lacks a consistent disaggregated rainfall values’ positioning and leads to significant spatial discontinuities between original EPS rainfall field’s pixels. Research is under way to increase this method’s usefulness. References  Olsson, J., et Lindström, G Evaluation and calibration of operational hydrological ensemble forecasts in Sweden. Journal of Hydrology, 350: 14– 24.  Perica, S., and Foufoula-Georgiou, E Model for multiscale disaggregation of spatial rainfall based on coupling meteorological and scaling descriptions. Journal Of Geophysical Research, 101(D21):  Peterson, W.W, Birdsall, T.G., and Fox, W.C., The theory of signal detectability. Trans. IRE prof. Group. Inf. Theory, PGIT, 2-4:  Rezacova, D., Sokol, Z., Pesice, P A radar-based verification of precipitation forecast for local convective storms. Atmospheric Research, 83: 211–224.